Understanding hidden AI in everyday business tools

Hidden AI is artificial intelligence developed and owned by a company and built into their everyday products and services. Its decisions and data access often operate entirely beneath the surface.

What I do: I find where AI is operating inside the everyday tools a business already uses and work out what information it's drawing on and what it can recommend, communicate or change as a result.
How I do it:I publish independent, plain-English explanations to help businesses understand these hidden capabilities.

Laptop on a wooden office desk displaying an inbox application alongside a smartphone and notebook, with colleagues discussing work in a bright plant-filled background.

Desk setup with an open laptop and external monitor displaying a four-person video conference call positioned near a window in a bright workspace.

Email, Work & Devices

What it does: Hidden AI in email clients, office suites, meeting assistants and workplace chats summarises correspondence, transcribes conversations and organises files across cloud storage.
What I examine: I look at what information these features access beneath the surface, whether default settings leave training switched on and what evidence is left behind when dealing with automated outcomes.
Did you know: Excluding a source from Google Gemini's searches does not stop it from using that content? Explicitly referenced or currently open files can still be processed that means "don't search these files" is very different from "don't process these files".

Social Media & Messaging

What it does: Background AI on social networks ranks feeds and targets advertising, while business assistants and messaging bots engage directly with customers, recommend products and handle routine inquiries.
What I examine: I review default platform training terms, regional objection routes and the boundaries between ordinary private communication and information submitted to AI.
Did you know: Handing a difficult customer conversation over to a human on WhatsApp doesn't necessarily create an information boundary? Meta Business Agent can still learn from conversations even after a human takes over.

Shopkeeper behind a counter handing a wrapped package to a customer next to a digital point-of-sale tablet showing customer communications.

Sales, Marketing & Customer Service

What it does: CRM platforms, helpdesk tools, and e-commerce software use built-in intelligence to score leads, write customer service responses, and interact directly with store data and customer transactions.
What I examine: I break down documented defaults, permissions, and whether AI features can change records or execute actions autonomously, helping businesses retain oversight of their operations.
Did you know: An app created by Shopify’s AI outlives the chat window that made it? A Sidekick-created app operates as an installed application in its own right, meaning ending the chat does not stop it, it requires separate review and removal.

Desk workspace with a laptop, notebook and magnifying glass showing diagrams connecting AI icons to EU GDPR data protection and padlock graphics.

Privacy & GDPR Exposure

What happens to privacy: Everyday business services often turn AI processing and training on by default, reading emails, customer records and connected information to share with outside companies or cross organisational boundaries.
What I examine: I identify where these hidden features create GDPR risks, such as missing privacy records and unverified assumptions and explain which settings require immediate attention.
Did you know: A read-only AI connection can still create unexpected data disclosures? Even if an AI tool cannot change your original files in Gmail, Drive or HubSpot retrieved information can still end up embedded inside a new output or design for a different audience.